A prompt library: the exact prompts behind eight NOLGIA examples
Good prompts are mostly structure. Here are eight that worked, exactly as they were sent, and the lines in each that do the heavy lifting.
- 8 min read
- Presets: Film scene, Car chase game trailer, Exploded Engine Showcase, Recreate a launch video, Logo reveal, Animated titles, Build a solar system, Digital Concept Art
These are the prompts behind eight examples on NOLGIA's preset pages, copied from the renders themselves, with the model each one ran on. Copy one, change the nouns, and keep the structure: that is where most of the work is.
Seven are video and one is a still. None of them names a real brand, a real person or a real place.
How to read these prompts
Every prompt on this page is the text that was sent to the model for a real NOLGIA example, copied from the render itself. The strongest ones follow the same order:
- The shot and the format first: what kind of shot, the aspect ratio, one take or several.
- The subject, in physical detail: materials, colours, what it is doing.
- The camera: where it starts, how it moves, where it ends.
- The light and the palette: where the light comes from and what colour it is.
- Physics and sound: how things should move and what you should hear.
- What must not appear: text, logos, signage, extra objects.
Tip:The last sentence about brands. Most of these prompts end with a sentence about brand names and logos. The preset that wrote the prompt added it, to keep real brands out of the frame. Keep it when you reuse a prompt.
A night chase in one take

One continuous action shot, 16:9, photoreal. SCENE: a courier on a black motorcycle threads through slow night traffic on a rain slicked elevated expressway, weaving BETWEEN two cars, sliding past a truck close enough to touch, then cutting hard right down an exit ramp into darkness, spray kicking off the rear wheel. Plain concrete barriers and dark glass towers, no billboards, no signage, no storefronts, no lettering on any surface. CAMERA: low tracking shot alongside the bike, speed felt through passing lights, then swinging behind it into the alley. LIGHTING: wet asphalt reflecting headlights and cool sodium lamps, beams cutting through drizzle. PHYSICS: real motorcycle lean and countersteer, believable traffic speeds, water behaves like water. Helmet stays on. Engine, rain and city ambience, no music. No on-screen text, no watermark.
What does the work:
- Labelled sections. SCENE, CAMERA, LIGHTING and PHYSICS each get their own sentence, so nothing competes for the same words.
- Real-world physics, named. "Real motorcycle lean and countersteer" is what keeps the bike from floating.
- What must not appear. No billboards, no signage, no lettering, which keeps the city generic and the frame clean.
To have the preset write a prompt like this from a one-line brief, open Film scene. To run this one yourself, pick Seedance 2.5 and paste it.
A dawn chase across a frozen lake

Game announce trailer shot, 16:9, one continuous in-world take. A stylised dawn chase across a FROZEN LAKE ringed by dense pine forest, with no buildings, roads, signs or structures of any kind anywhere: a wide sheet of blue grey ice under thin snow, black pines on the shore, low mist and a pale pink sunrise. An invented low wide black coupe with a long bonnet, a single thin horizontal tail light and smooth badge free panels drifts across the ice throwing long rooster tails of snow, two invented pursuit cars of no real make, low wedge shaped coupes in plain matte white with no grille at all, a single unbroken horizontal light strip across the nose and another across the tail, smooth unbroken panels, plain roofs with no light bars, no badges, no emblems, no stripes, no livery and no markings of any kind, blank plate recesses, fishtailing in behind it. The camera runs alongside at bumper height, drops back through the pursuit pack, then rises to just above roof height and pushes forward over the pursuit pack, holding the black coupe large in the centre of frame as it powers away toward the far shore, with the pine treeline staying low in the background the whole time. All the colour comes from the sky and the ice: pink and gold sunrise on the snow, cold blue shadow, headlights cutting the mist. Stylised game cinematic look: heightened contrast, saturated reflections on the ice, a little motion blur on the surface, clean simplified geometry with rich materials rather than photoreal grit. Snow spray in the headlight beams, ice chips where a wheel bites. The vehicles are unbranded, carrying no logo, wordmark, number, symbol or trademark of any kind, every panel plain, and their number plates are blank. Engines, tyres scrabbling on ice, wind and a low pulsing bass, no vocals. No on-screen text, no watermark. Use only the brand names, logos and trademarks the customer includes in their request; never add real-world brands, logos or trademarked designs the customer did not mention.
What does the work:
- The camera path, in order. Alongside, then back through the pack, then up and forward. The model follows a sequence better than a single adjective.
- Where the colour comes from. "All the colour comes from the sky and the ice" stops the model inventing neon.
- Invented vehicles, described to the panel. No badges, no grille, blank plates: the look of a car without a real make.
To have the preset write a prompt like this from a one-line brief, open Car chase game trailer. To run this one yourself, pick Seedance 2.5 and paste it.
An exploded watch movement

Design studio product film, 16:9, one continuous slow orbit. A mechanical wristwatch movement hangs in a soft dark grey void, exploded along its central axis: every part suspended in order on an invisible spine with clean even spacing, a plain smooth case back, a brushed steel mainplate, two gilt bridges, a train of tiny brass gear wheels, a balance wheel with its coiled blue hairspring, a crown on its stem, a plain blank dial disc with a smooth unmarked face and a domed sapphire crystal. The camera orbits smoothly through a quarter turn while the parts drift a little further apart, then ease gently back toward each other. Studio product lighting, one soft key from high left, a cool rim light picking out every polished edge, soft reflections on a seamless dark floor, crisp macro detail on the gear teeth and screw heads. The movement is unbranded, carrying no logo, wordmark or trademark of any kind, and every surface including the dial is clean and unlettered. Quiet room tone with faint metal ticks as the parts settle, no music. No on-screen text, no watermark. Use only the brand names, logos and trademarks the customer includes in their request; never add real-world brands, logos or trademarked designs the customer did not mention.
What does the work:
- Every part, in assembly order. The list is the storyboard: case back, mainplate, bridges, gears, balance wheel, crystal.
- One motion, with a return. Drift apart, then ease back. A clip with a beginning and an end cuts into an edit more easily.
- Studio light, spelled out. One soft key from high left and a cool rim light, the vocabulary of a real product shoot.
To have the preset write a prompt like this from a one-line brief, open Exploded Engine Showcase. To run this one yourself, pick Seedance 2.5 and paste it.
A product hero shot in the dark

Product launch film hero shot, 16:9, one continuous take. A short cylindrical smart speaker wrapped in dark green fabric mesh with a matte black top disc and a thin brass ring at its base stands on a low round black plinth in a dark wide studio. The camera starts almost directly above looking down on the top disc and descends in a slow arc, settling into a three quarter view at the speaker's own height, while a single hard amber rim light from the back right traces the brass ring and the weave of the mesh and a low soft fill from the front lifts the fabric out of the dark so the texture, the top disc and the plinth all read clearly. Behind it the studio falls away into a soft deep green gradient rather than pure black, with one faint warm pool of light on the floor for depth. Fingerprint clean, deliberate and expensive, the speaker large in frame with room above and below it. The speaker is unbranded, carrying no logo, wordmark, engraving, light pattern or trademark of any kind, and every surface is clean and unlettered so the titles can be laid on afterwards. Quiet studio room tone with a faint electrical hum from the lighting, no music and no voice. No on-screen text, no watermark. Use only the brand names, logos and trademarks the customer includes in their request; never add real-world brands, logos or trademarked designs the customer did not mention.
What does the work:
- Materials the light can find. Fabric mesh, a matte top disc, a thin brass ring: each one gives the rim light something to trace.
- Room left for titles. The product is large in frame with space above and below, and every surface is unlettered so type can go on later.
- A background that is not black. The prompt asks for a soft deep green gradient, with one warm pool of light on the floor for depth.
To have the preset write a prompt like this from a one-line brief, open Recreate a launch video. To run this one yourself, pick Seedance 2.5 and paste it.
A logo sting with a specular sweep

A logo reveal on a deep charcoal seamless ground, 16:9, premium motion design finish. A bold abstract mark builds itself at the centre of frame: three thick rounded chevrons, one in polished warm brass, one in deep teal matte and one in brushed off white, arriving from off screen on smooth eased arcs, rotating into alignment and locking together into a single compact geometric mark with one soft settle, a bright specular highlight sweeping across the brass as the last piece lands. The mark then holds still and centred, filling about a third of the frame height, while a warm circular light bloom breathes behind it and fine dust motes drift through the key light. Studio three point lighting, polished brass and soft matte finishes with crisp bevelled edges, shallow depth of field, elegant restrained motion with no bounce and no overshoot. The mark is a pure abstract shape with no letters, no words and no numbers anywhere in the frame. A soft low whoosh into one warm settle tone, no music. No on-screen text, no watermark. Use only the brand names, logos and trademarks the customer includes in their request; never add real-world brands, logos or trademarked designs the customer did not mention.
What does the work:
- Three materials, three pieces. Brass, matte teal and brushed off white keep each piece readable as it moves.
- Motion rules. "No bounce and no overshoot" is the line that makes it feel restrained rather than cartoonish.
- Hold at the end. The mark stays still and centred, so the clip ends on something you can cut away from.
To have the preset write a prompt like this from a one-line brief, open Logo reveal. To run this one yourself, pick Seedance 2.5 and paste it.
A title card that spells its words

A single animated title card in a cinematic trailer style, 16:9, on a pure flat black field with nothing else in frame. Polished brushed chrome capital letters with sharp bevelled edges and a warm specular highlight sweeping across them, reading exactly and only: THE LONG WAY HOME. The words are spelled letter for letter T H E space L O N G space W A Y space H O M E, centred, wide letterspacing, one line, correctly formed english letterforms, no misspelling, no extra letters, no duplicated words. Very slow push in on the type over the whole clip, a faint anamorphic light streak crossing the letters once. No other text anywhere, no subtitle, no credit block, no logo, no watermark, no people, no objects, no background scene. Deep blacks, crisp highlights, a low cinematic drone in the audio, no dialogue.
What does the work:
- Spell the words twice. Once as words and once letter by letter, which is how the Animated titles preset asks for exact type.
- Say what else must not appear. No subtitle, no credit block, no logo: otherwise models like to add small print.
- One move, the whole clip. A slow push in over the full length, and one light streak, once.
To have the preset write a prompt like this from a one-line brief, open Animated titles. To run this one yourself, pick Gemini Omni 1.1 Flash and paste it.
A ringed planet and its moons

An astronomy visualization, 16:9, photoreal scientific rendering against deep black space with a faint dust of distant stars. The camera drifts slowly outward from a warm ringed gas giant whose banded cloud belts turn visibly and whose ring shadow falls across the southern hemisphere, while three moons track along their orbital paths at different speeds, each drawing a faint luminous trail behind it. As the camera pulls back the whole system comes into frame and the orbital ellipses resolve as thin glowing lines on the plane of the system. Hard directional sunlight from screen left with a true terminator across every body, realistic surface detail, subtle atmospheric limb glow, no lens flares, no volumetric god rays. Calm and precise, one continuous move with no cuts. Nothing is labelled and there is no lettering, no numbers and no on-screen text anywhere. Deep low space ambience, no music, no narration. No watermark. Use only the brand names, logos and trademarks the customer includes in their request; never add real-world brands, logos or trademarked designs the customer did not mention.
What does the work:
- Physically true light. Hard sunlight from one side and "a true terminator across every body" is what makes it look like a real planet.
- What to leave out. No lens flares, no god rays: the effects that make space footage look fake.
- Nothing labelled. Diagram-style prompts invite captions and numbers, so the prompt rules them out.
To have the preset write a prompt like this from a one-line brief, open Build a solar system. To run this one yourself, pick Seedance 2.5 and paste it.
A creature key frame for an art department

A digital concept art key frame painted for a film art department, 16:9, a creature design study set in its habitat. A vast six legged grazing beast, the size of a house, crosses a frozen shallow lake at first light, its back armoured in overlapping plates furred with pale moss, a low heavy head with wide flat teeth for stripping bark, breath steaming in the cold. Two small figures in fur hoods stand beside a sledge on the ice in the foreground, tiny against the animal, carrying the scale. Believable anatomy with weight in every step, the front legs planted and the hind pair mid stride, ice cracking under the nearest foot. Painterly brushwork with visible strokes in the foreground ice and softer atmospheric handling into the distant pine shore, a limited palette of cold slate grey, pale bone and dull moss green with one warm rose accent in the dawn sky, strong aerial perspective, cinematic value structure with the darkest notes held in the creature's underbelly. Matte painting finish, no photography, no readable lettering anywhere, no signage. Use only the brand names, logos and trademarks the customer includes in their request; never add real-world brands, logos or trademarked designs the customer did not mention.
What does the work:
- Scale, carried by people. Two small figures beside a sledge tell you the creature is the size of a house.
- Anatomy with a purpose. Flat teeth for stripping bark, weight in every step: the details a creature designer would ask about.
- A limited palette with one accent. Slate, bone and moss, and one warm rose in the sky to pull the eye.
To have the preset write a prompt like this from a one-line brief, open Digital Concept Art. To run this one yourself, pick GPT Image 2.5 Sunburst and paste it.
Make one of these yours
Keep the structure and change the nouns. Swap the courier for your own subject, the expressway for your location, the amber rim light for your palette, and leave the camera, physics and exclusion lines as they are.
- Change one section at a time, so you know which change moved the result.
- Keep the exclusions. They are the lines that keep stray text, logos and signage out of the frame.
- Want a different camera move? The video composer has a row of camera moves you can pick instead of typing one.
Watch for this:A prompt is not a guarantee. The same prompt on the same model gives a different take each time. These are the takes we kept.
What it costs
The prompts cost nothing to copy. A run is priced on the Generate button before you start, and a failed render is refunded. Live published rates for the three models on this page:
Seedance 2.5Pro and up
Per 5s clip
480p 26 credits · 720p 56 credits · 1080p 137 credits
1080p renders at 16:9 and 9:16
Gemini Omni 1.1 FlashPro and up
Per 5s clip
Standard 50 credits
GPT Image 2.5 SunburstEvery plan
Per image
Native 8 credits · 2K 21 credits · 4K 58 credits
Read live from the catalog when this page loads. Every price is shown before you generate. See every rate.
Questions and answers
- Are these the exact prompts?
- Yes. Each one is copied from the render it made, including the lines a preset added. The clip or image above it is that render.
- Will a prompt work on a different model?
- Usually, with a different take. The structure carries over; details like how a model reads text or camera language vary, so expect to run it more than once.
- Why are they so long?
- Because every line closes a door the model would otherwise walk through: extra text, the wrong light, a floating vehicle. A short prompt works too, but the model fills the gaps with its own choices.
- Do I have to write prompts like this?
- No. Each preset on this page takes a one-line brief and writes a prompt like these for you.
Take a prompt and make it yours
Open the video composerRelated guides
Video editing
Keep the subject, change the world around it
Keep the legs, hands or product in a clip and redraw the place around them. The prompt order that held in 6 of 6 tests, and where it breaks.
· 9 min read
Blog · Model
Seedance 2.5 at 1080p, with reference, edit and extend
Seedance 2.5 renders 1080p in 16:9 and 9:16, takes up to ten reference videos and ten audio tracks, and can reference, edit or extend a clip of yours.
· 5 min read
Blog · Comparison
One prompt, six video models, nine measured checks
One prompt with nine checkable details, sent to six AI video models twice each, and every clip measured with a detector, text reading and camera tracking.
· 12 min read




